The Risk Measurement Used The High Frequency Data Of CSI300 Index Based On The Realized-GARCH Model

Since J.P.Morgan company developed the Value-at-risk method.,VaR has become widely used in the risk management as one criteria of measurement,it provides a uniform method to the risk management.By this method,the potential losses of asset which is the most important aspect of risk management can be expressed via a monetary unit.This can describe the market risk of asset during a given period of time in a simple way.To calculate the VaR.there are lots of methods,such as Risk-Metrics,the quantitative methods using volatility model,empirical distribution method and extreme value method.The method that uses the volatility model,such as GARCH model,is still the main method.Using GARCH models to model the volatility dynamically can describe the characteristics of dynamic asset return changes in a better way.The traditional research usually focuses on the low frequency data.In the financial markets,especially in the stock market,information has a continuous effect on the changing process of asset returns,so low frequency data may cause the different level losses of the information.With the development of the technology and the availability of data,how to use high-frequency data into researching the volatility of the financial markets has become a hot issue in the field of financial research.High frequency data is securities exchange data which has shorter time interval,and this interval usually less than one day.In view of the high frequency data,there is a new method to measure the volatility—the realized volatility.The realized volatility can make the full use the daily information when predicting the future volatility.What is more,it is the unbiased and effective estimate of the day volatility.In view of the advantages of the realized volatility,more and more research try to combine the realized volatility with the GARCH models and thus build a model of fluctuations.Except the GARCH-X model which add the realized volatility into the GARCH models as an exogenous variable directly,in 2012,Hansen invited a model which is called the realized-GARCH model.He also added the realized volatility into the GARCH model.The realized-GARCH model is a complete model.This model added the realized measure into the traditional GARCH models,it use a formula which is called "measurement equation",thisformula combined the realized volatility and conditional variance.As a result,this new model integrate the realized measure and the conditional variance in one complete frame.Under the background talked above,this paper has done much research on the Realized-GARCH model,including the empirical research and the extended applications.And in order to provide a better reference to the risk management.For example,to choose a more suitable model and more accurate error distribution for the estimation of VaR.VaR and its calculation method are introduced in this paper,especially focus on the parametric method based on the volatility models.And then get the importance of the volatility modeling.Made the summarize of the volatility models including the ARCH and the GARCH models.By analyzing the characteristic of high frequency data,the Realized-GARCH model that combined with the realized volatility has been fully analyzed in this paper.On the basis of theoretical analysis,this paper used two periods of 5 minutes-high frequency data from China stock market.To the empirical analysis which include the model estimation,data fitting,volatility forecasting and VaR estimation and so on.What's more,the empirical results of two periods are similar,which reinforces the conclusion of this paper.In addition,in order to fit the features of financial high frequency data such as leptokurtosis and fat-tail,this paper extents Realized-GARCH model proposed to make its residual error follow fat-tail distribution,such as t-distribution,Skewed-t distribution and GED distribution,and then make adequate comparison analysis.Results show that this kind of promotion can significantly enhance the model's fitting and forecast capability,which can improve the effect of risk measurement,and has big significance to improve risk management ability.Through empirical researching,this paper got the following conclusion:In the aspect of volatility modeling,Realized-GARCH model can reflect the fat-tail features of financial data better compared with the traditional GARCH models.Under this model,the parameter estimation is more stable,the fitting degree is higher and the volatility forecast is more accurate.Furthermore,from the model fitting and volatility forecasting,the Realized-GARCH model based on GED distribution had the optimal performance.In terms of risk measurement,through the calculation and evaluation of VaR,the results also show that Realized-GARCH model is better than GARCH model,and it can get tail risk more accurately.Under the confidence level of 1%and 5%,The Realized-GARCH model based on the fat-tailed distribution has a more precise forecast on VaR.Finally,after the comprehensive analysis of the results above,this paper make the conclusion that Realized-GARCH model based on GED distribution is a more suitable mode in volatility modeling using high frequency data and risk measurement.